Anthropic is circulating a draft AI safety bill in Washington, according to a Politico report cited by TechCrunch. The proposal would let the government block release of an AI model in scenarios involving a ‘foreseeable, imminent risk’: more than 50 people at risk of dying or more than $1 billion in property damage. The harm must stem from the creation or release of a biological weapon, the model evading its developer’s controls, or a cyberattack. Those thresholds match a California law the article names as Senate Bill 53, which uses the same figures to define ‘catastrophic risk.’

The enforcement mechanism depends on ‘independent evaluators’ registered with the U.S. Commerce Department. AI labs could pay the evaluators as long as payment is not contingent on the evaluation’s results, though the article notes it is unclear how that prevents payments from influencing outcomes. The draft also arrives amid proposals such as DeepMind chairman Demis Hassabis’s suggestion that a self-governing AI regulatory body should mimic FINRA. Under the bill, an evaluator who identifies an imminent catastrophic risk would have 72 hours to contact Commerce, and Commerce would then have to request that a D.C. judge halt the model’s release; no federal agency could block distribution on its own discretion.

An Anthropic spokesperson confirmed the draft to Politico and said it reflects the proposals in Anthropic‘s Advanced AI Framework, which the company has advocated with federal and state lawmakers since June. The article points out that these death-and-damage scenarios are somewhat removed from the industry’s current top concern: agentic hacking, in which AI programs disobey their operators and attack websites. Labs like Anthropic have been offering their own preferred solution, voluntary embedded evaluators positioned as neutral third-party observers inside AI organizations.

Anthropic‘s legislative push goes further than some industry positions; Nvidia CEO Jensen Huang has argued the AI industry needs no new laws. TechCrunch also notes that despite labs’ recurring warnings about civilizational risk, they have largely resisted mandatory government oversight.